• Critical Realignments were once a common event in American politics

• A major change in the party system occurred about 30 years until 1936

• Beginning in 1936, the party system froze outside the south: Democrats party of labor, Republicans party of business (Lipset & Rokkan, 1967)

• One impediment to broad change: stable partisanship attachments

• In 21st century elections, 90% of Republicans support their party's Presidential candidate

• During Watergate, majority of Republicans supported Nixon (dipped below 50% twice)

• Around 80% of Republicans approve of President Trump's job performance

• "However party and religious identifications come about, once they take root in early adulthood, they often persist. Partisan identities are enduring features of citizens’ self-conceptions." (Green, Palmquist & Schickler, 2002)

• American political behavior is usually assumed to be time-invariant

• "Extraordinary durability of voting habits" (Key, 1949)

• Implicit in idea of "The Normal Vote" (Converse, 1964)

• "[Absent realignments], as with religion, people are often adherents of a particular political party because their great-grandparents favored it for entirely different reasons" (Achen & Bartels, 2016)

• Between 2012 and 2017, 13% of partisans (net) switched from one major party to the other (Voter Study Group)

• In the Pew American Trends panel, 10% of partisans (net) switched

• Switching was concentrated among the young, age was demographic most predictive of switching

• Show that Partisan Defections were higher Pre-realignment

• Show that demographics weren't predictive of partisanship

• Evaluate whether partisanship is unstable broadly or only among new entrants to the electorate

• First individual-level dataset covering period before the New Deal

• High levels of partisan instability were the norm from 1908 to 1928

• Social groupings don't explain partisanship (important for theoretical predictions see Mason (2016))

• Classic theories of socialization don't explain heterogeneities in party stability before 1930

• Current instability is structurally different than instability before the New Deal

• California Great Registers, 1908-1968 (today through 1930 only)

• Registers contains 57 million name, address, occupation and party registration records

• California had closed primaries

• Microfilm copies scanned and converted to text by Ancestry.com

• Supplemented with personally identified 1920-1940 census data from the Minnesota Population Center (IPUMS), matched exactly on name and address

• California Great Registers, 1908-1930

• Registers contains 57 million name, address, occupation and party registration records

• California had closed primaries

• Microfilm copies scanned and converted to text by Ancestry.com

• Supplemented with personally identified 1920-1930 census data from the Minnesota Population Center (IPUMS), matched exactly on name and address

• Only state where clean individual panel data is available at scale

• We know the New Deal realignment happened there, can parse mechanism

• Except for 1924, county-level Presidential vote share similar to rest of non-South

• Literary Digest Polls show Presidential vote switching only 3-5 points higher in California than rest of country (Erikson & Tedin, 1981)

• Predates scientifically sampled polls

• Window into political behavior under a different party system

• Issues contested were very different

• Post-1936 U.S. party system froze around issues of capital vs labor (Lippset & Rokkan, 1967)

• Maybe freezing of partisanship and axis of policy contestation wasn't mere coincidence

• Group identities were weak, little difference in partisanship by age, gender (See Spahn's Before The American Voter, 2018)

• Small differences by parents' national origin, class

• 2014 - 2016 (major party voters): 10% (Pew, 2017)

• 1992-1996 (major party voters): 4% (ANES, 1997)

• 1965-1973 (major party voters): 6% (Jennings & Niemi, 1973)

• 1908-1928 (major party voters): 10% every four years

• 1920-1930 (major party voters): 16%

• Match Great Registers to themselves in successive Presidential Years

• Match exactly on Name and Address

• Errs strongly on the side of caution

• Matching errors lead to inflated instability, so minimize instability by matching exactly

• Restricting to non-movers reduces switching rate by 15%

• Exact matching gives nearly identical results to Bayesian probabilistic record linkage (McVeigh, Spahn & Murray, 2018)

Need to test if high rate of party switching is compositional or broad-based. Socialization theory says…

• People with more experience voting for a party should be more stable (partisan habituation)

• Older people should be more stable (longer period of socialization)

• Voters just coming of age should be especially unstable (generational model, Ghitza & Gelman, 2014)

• California Women first voted in 1912

• Claim: because parents didn't expect them to vote, would have gotten less partisan socialization from parents (Andersen, 1979)

• Conventional prediction: less partisan socialization would lead to more instability, voters "not immunized" to political forces by childhood partisan socialization

• Claim in literature: Women contributed to new deal realignment by converting to Democratic Party more often than men (Corder & Wolbrecht, 2016)

• Women's registration rate about 75% of men's

• Matching process excludes women that change their name

Party Switching Rates for Native-Born Americans:

• 10% (1916 - 1920)
• 14% (1920 - 1924)

Party Switching Rates for Foreign-Born Americans:

• 14% (1916 - 1920)
• 18% (1920 - 1924)

• Except for modest age effects, little heterogeneity in switching pre-realignment

• Socialization is less important in partisan attchments pre-1930

• Perhaps most and least experienced voters switch parties for similar reasons

• Non-socialized attachments may be in play

• The fact that identities aren't differentiated in partisan may contribute to weak attachments (Mason, 2016)

• Generational Replacement has replaced realignment as the engine of political change

• High rates of switching by young people is consistent with the formation of political generations (Ghitza & Gelman, 2014)

• Before 1930, little partisan differentiation by age

• Present day: older voters are most Republican, declining as share of voters

• Present day: youngest voters are most Democratic, rising as share of electorate

• New data makes historical political behavior tractable

• Individual partisanship less stable before 1928

• Instability was hidden by stable macropartisanship

• Present era of partisan stability began after New Deal realignment

• Folk theory doesn't hold pre-realignment

• Present instability is different than pre-realignment, concentrated among the young

• Barring broad instability, critical realignment seems unlikely

• Family socialization and gender dynamics with Shanto Iyengar

• 100 years of partisan geographic sorting with Jonathan Rodden & Shanto Iyengar

• Decompose the New Deal realignment into component sources of change: conversion, mobilization, migration (conversion accounts for almost all of it)

• Match 1920 and 1930 censuses exactly to Great Registers of same year

• Use fastlink (Enamorado et al) to match 1920 to 1930 census

• Match censuses on first, last, street name, age, occupation and place of birth of mother, father and self

• Use only matches with probability of matching > .995

• In 1930, voters passed referendum amending registration laws

• Starting in 1932 election, those that voted in any election had registration rolled over to following election

• Non-voters (no vote in general or primary) were still purged

• To isolate new registrants, identify voters on Presidential rolls, but missing from midterm rolls

• Use fastlink from Enamorado, Fifield & Imai to match extremely loosely, identify non-voters

• Cases that had at least one match of Pr > .2 considered matched. Allows many false positives to minimize false negatives.

• Match rate post-1936: 88.6%